A new method for predicting binding affinity in computer-aided drug design

A new semi–empirical method for calculating free energiesof binding from molecular dynamics (MD) simulations is presented.It is based on standard thermodynamic cycles and on a linearapproximation of polar and non–polar free energy contributionsfrom the corresponding MD averages. The method is tested ona set of endothiapepsin inhibitors and found to give accurateresults both for absolute as well as relative free energies.     

Binding free energy calculations for P450cam-substrate complexes

A recently proposed semi-empirical method for calculating bindingfree energies was used to examine the binding of a variety ofsubstrates to cytochrome P450cam. For a set of 11 differentpotential substrates of cytochrome P450cam, both the absoluteand relative binding free energies were generally well reproduced.The mean error in the calculated absolute binding free energyfor all 11 compounds is 0.55 kcal/mol. Forty-eight out of 55calculated relative binding free energies have the correct signand the mean unsigned error between calculated and experimentalrelative binding free energies is 0.77 kcal/mol. For one substrate,thiocamphor, the effect of substrate orientation on the calculatedbinding free energy was examined. The ability of this methodto predict the effect of active site mutations was also examinedin two cases.     

Molecular mechanics analysis of drug-resistant mutants of HIV protease

Drug-resistant mutants of HIV-1 protease limit the long-termeffectiveness of current anti-viral therapy. In order to studydrug resistance, the wild-type HIV-1 protease and the mutantsR8Q, V32I, M46I, V82A, V82I, V82F, I84V, V32I/I84V and M46I/I84Vwere modeled with the inhibitors saquinavir and indinavir usingthe program AMMP. A new screen term was introduced to reproducemore correctly the electron distribution of atoms. The atomicpartial charge was represented as a delocalized charge distributioninstead of a point charge. The calculated protease–saquinavirinteraction energies showed the highly significant correlationof 0.79 with free energy differences derived from the measuredinhibition constants for all 10 models. Three different protonationstates of indinavir were evaluated. The best indinavir modelincluded a sulfate and gave a correlation coefficient of 0.68between the calculated interaction energies and free energiesfrom inhibition constants for nine models. The exception wasR8Q with indinavir, probably due to differences in the solvationenergy. No significant correlation was found using the standardmolecular mechanics terms. The incorporation of the new screencorrection resulted in better prediction of the effects of inhibitorson resistant protease variants and has potential for selectingmore effective inhibitors for resistant virus.     

Relative binding free energy calculations of inhibitors to two mutants (Glu46-- Ala/Gln) of ribonuclease T1 using molecular dynamics/free energy perturbation approaches

We present free energy perturbation calculations on the complexesof Glu46— Ala46 (E46A) and Glu46— Gln46 (E46Q) mutantsof ribonuclease T1 (RNaseT1) with inhibitors 2‘-guanosinemonophosphate (GMP) and 2’adenosine monophosphate (AMP)by a thermodynamic perturbation method implemented with moleculardynamics (MD). Using the available crystal structure of theRNaseT1–GMP complex, the structures of E46A-GMP and E46Q-GMPwere model built and equilibrated with MD simulations. The structuresof E46A-AMP and E46Q-AMP were obtained as a final structureof the GMP—AMP perturbation calculation respectively.The calculated difference in the free energy of binding (Gbind)was 0.31 kcal/mol for the E46A system and —1.04 kcal/molfor the E46Q system. The resultant free energies are much smallerthan the experimental and calculated value of 3 kcal/mol forthe native RNase T1, which suggests that both mutants have greaterrelative adenine affinities than native RNaseT1. EspeciallyE46Q is calculated to have a larger affinity for adenine thanguanine, as we suggested previously from the calculation onthe native RNaseT1. Thus, the molecular dynamics/free energyperturbation method may be helpful in protein engineering, directedtoward increasing or changing the substrate specificity of enzymes.     

Molecular mechanics analysis of inhibitor binding to HIV-1 protease

Crystallographic structures of HIV protease with three differentpeptide-mimetic inhibitors were subjected to energy minimizationusing molecular mechanics, the minimized structures analyzedand the inhibitor binding energies calculated. Partial chargeassignment for the hydrogen bonded catalytic aspartk acids,Asp25 and -25', was in good agreement with charge calculationsusing semi-empirical molecular orbital methods. Root mean squaredeviations on minimization were small and similar for both subunitsin the protease dimer. The surface loops, which had the largestB factors, changed most on minimization; the hydrophobic coreand the inhibitor binding site showed little change. The distance-dependentdielectric of D(r) = 4r was found to be preferable to D(r) =r. Distance restraints were applied for the intermolecular hydrogenbonds to maintain the conformation of the inhibitor bindingsite. Using the dielectric of D(r) = 4r, the calculated interactionenergy of the three inhibitors with the protease ranged from–53 to –56 kcal/mol. The groups of the inhibitorswere changed to add or remove a ‘transition state analogue’hydroxyl group, and the loss in energy on the removal of thisgroup was calculated to be 0.9–1.7 kcal/mol. This wouldrepresent 19–36% of the total measured difference in bindingenergy between the inhibitors JG365 and MVT-101.     

The nature of the ion binding interactions in EF-hand peptide analogs: free energy simulation of Asp to Asn mutations

The binding of the La³⁺ ion to a tridecapeptide, which is amodel for the EF-hand in calcium-binding proteins, is studiedhi solution by free energy simulations. The calculations analyzethe effect on the La³⁺ ion binding of the mutation of Asp toAsn for side chains that interact directly with the ion. Theresults are compared with the measurements of Marsden.B.J.,Hodges, R.S. and Sykes, B.D. (1989) Biochemistry, 28,8839, onthe same system. They found that the Asp to Asn mutation hasonly a small effect on the binding; the observed differencesin the free energies on changing one Asp to an Asn are between-0.3 and 1.8 kcal/ mol. This result is analyzed by alchemicalsimulations for the tridecapeptide in the bound Qoop) structureand free (extended) form. The free energy changes due to themutation of an Asp to an Asn are large and positive for boththe bound and free forms. However, since the values of the freeenergy changes are calculated to be similar hi the two forms,the difference in the binding free energy of Asp and Asn peptidesis found to be small, in agreement with experiment. By use ofthermodynamic integration, the various contributions to thefree energy changes are estimated. In the com-plexed form, theAsp to Asn mutation is favored by the reduction in the repulsiveinteraction with other charged residues of the peptide; it isdisfavored by the reduction of the stabilization of the ionand the surrounding water has a small effect. When the peptideadopts an extended conformation in the absence of the ion, themutation Asp to Asn is strongly disfavored by the interactionswith the water and is favored by the interactions within thepeptide. The results demonstrate the essential role of contributionsto the binding of EF-hands from interactions other than thosebetween the ion and the charged amino acid side chains. Theresults obtained from the simulations suggest, in accord withcrystal structures of La³⁺ bound to various ligands, that thecalcium-binding loop complexed with La³⁺ in solution has a significantlydifferent structure from that observed hi proteins.     

The Fourier-Green's function and the rapid evaluation of molecular potentials

Two tasks must be accomplished when calculating the bindingmodalities and binding energies of two molecules in solution:the calculation of the interaction energy and the calculationof the effects of solvation. It is the competition between theenergy of binding and the energy of remaining solvation whichdetermines the binding properties. It is necessary to calculate(or at least approximate in some manner) the partition functionin order to make a theoretical estimate of these effects. Anefficient algorithm for performing the energy evaluations necessaryfor this calculation is presented in this paper. The fast Fouriertransform (FFT) is used in combination with a polar factorizationof the potentials to calculate the interaction energy at allrelative translations between two molecules of fixed orientation.Thermodynamic quantities, including the partition function,internal and free energies can then be estimated from a setof these calculations covering the orientation space.     

Free energy perturbation studies on binding of A-74704 and its diester analog to HIV-1 protease

Free energy simulations have been employed to rationalize thebinding differences between A-74704, a pseudo C₂- symmetricinhibitor of HIV-1 protease and its diester analog. The diesteranalog inhibitor, which misses two hydrogen bonds with the enzymeactive site, is surprisingly only 10-fold weaker. The calculatedfree energy difference of 1.7 ± 0.6 kcal/mol is in agreementwith the experimental result. Further, the simulations showthat such a small difference in binding free energies is dueto (1) weaker hydrogen bond interactions between the two (P₁and P₁) NH groups of A-74704 with Gly27/Gly27' carbonyls ofthe enzyme and (2) the higher desolvation free energy of A-74704compared with its ester analog. The results of these calculationsand their implications for design of HIV-1 protease inhibitorsare discussed.     

The stability and unfolding of an IgG binding protein based upon the B domain of protein A from Staphylococcus aureus probed by tryptophan substitution and fluorescence spectroscopy

The stability and unfolding of an immunoglobulin (Ig) G bindingprotein based upon the B domain of protein A (SpAB) from Staphylococcusaureus were studied by substituting tryptophan residues at strategiclocations within each of the three a-helical regions (al-a3)of the domain. The role of the C-terminal helix, a3, was investigatedby generating two protein constructs, one corresponding to thecomplete SpAB, the other lacking a part of ct3; the Trp substitutionswere made in both one-and two-domain versions of each of theseconstructs. The fluorescence properties of each of the single-tryptophanmutants were studied in the native state and as a function ofguanidine-HCl-mediated unfolding, and their IgG binding activitieswere determined by a competitive enzyme-linked immunosorbentassay. The free energies of folding and of binding to IgG foreach mutant were compared with those for the native domains.The effect of each substitution upon the overall structure andupon the IgG binding interface was modelled by molecular graphicsand energy minimization. These studies indicate that (i) 3 contributesto the overall stability of the domain and to the formationof the IgG binding site in l and 2, and (ii) al unfolds first,followed by 2 and 3 together.     

Application of free energy simulations to the binding of a transition-state-analogue inhibitor to HTV protease

Free energy simulations (slow-change method) have been usedto estimate quantitatively the ratio of the binding constantsof (S) and (R) isomers of a novel HIV protease inhibitor, JG365.As a starting geometry, we used the X-ray crystallographic structureof a complex of HTV protease and JG365 provided by A.Wlodawer.According to our results the (S) configuration, i.e. the formpreviously identified experimentally, binds considerably moretightly to the protease (G° = 2.9 kcal/mol). When the (S)inhibitor is bound, there is a very strong preference for protonationof the Aspl25 (rather than the Asp25) residue of the protease.This study is the first to apply a new method for quantitativelyassessing the precision of free energies calculated by the slow-changemethod     

A three-dimensional construction of the active site (region 507749) of human neutral endopeptidase (EC.3.4.24.11)

A three-dimensional model of the 507–749 region of neutralendopeptidase-24.11 (NEP; E.C.3.4.24.11) was constructed integratingthe results of secondary structure predictions and sequencehomologies with the bacterial endopeptidase thermolysin. Additionaldata were extracted from the structure of two other metalloproteases,astacin and stromelysin. The resulting model accounts for themain biological properties of NEP and has been used to describethe environment close to the zinc atom defining the catalyticsite. The analysis of several thiol inhibitors, complexed inthe model active site, revealed the presence of a large hydrophobicpocket at the S₁' subsite level. This is supported by the natureof the constitutive amino acids. The computed energies of boundinhibitors correspond with the relative affinities of the stereoisomersof benzofused macrocycle derivatives of thiorphan. The modelcould be used to facilitate the design of new NEP inhibitors,as illustrated in the paper.     

Calculations of antibody-antigen interactions: microscopic and semi-microscopic evaluation of the free energies of binding of phosphorylcholine analogs to McPC603

The study of antibody-antigen interactions should greatly benefitfrom the development of quantitative models for the evaluationof binding free energies in proteins. The present work addressesthis challenge by considering the test case of the binding freeenergies of phosphorylcholine analogs to the murine myelomaprotein McPC603. This includes the evaluation of the differentialbinding energy as well as the absolute binding energies andtheir corresponding electrostatic contributions. Four differentapproaches are examined: the Protein Dipoles Langevin Dipoles(PDLD) method, the semi-microscopic PDLD (PDLD/S) method, afree energy perturbation (FEP) method based on an adiabaticcharging procedure and a linear response approximation thataccelerates the FEP calculation. The PDLD electrostatic calculationsare augmented by estimates of the relevant hydrophobic and stericcontributions. The determination of the hydrophobic energy involvesan approach which considers the modification of the effectivesurface area of the solute by local field effects. The stericcontributions are analyzed in terms of the corresponding reorganizationenergies. This treatment, which considers the protein as a harmonicsystem, views the steric forces as the restoring forces forthe electrostatic interactions. The FEP method is found to giveunreliable results with regular cut-off radii and starts togive quantitative results only in very expensive treatment withvery large cut-off radii. The PDLD and PDLD/S methods are muchfaster than the FEP approach and give reasonable results forboth the relative and absolute binding energies. The speed andsimplicity of the PDLD/S method make it an effective strategyfor interactive docking studies and indeed such an option isincorporated in the program MOLARIS. A component analysis ofthe different energy contributions of the FEP treatment anda similar PDLD analysis indicate that electrostatic effectsprovide the largest contribution to the differential bindingenergy, while the hydrophobic and steric contributions are muchsmaller. This finding lends further support to the idea thatelectrostatic interactions play a major role in determiningthe antigen specificity of McPC603.     

Empirical free energy calculations of ligand-protein crystallographic complexes. I. Knowledge-based ligand-protein interaction potentials applied to the prediction of human immunodeficiency virus 1 protease binding affinity

The steadily increasing number of high-resolution human immunodeficiencyvirus (HIV) 1 protease complexes has been the impetus for theelaboration of knowledge-based mean field ligand-protein interactionpotentials. These potentials have been linked with the hydrophobicityand conformational entropy scales developed originally to explainprotein folding and stability. Empirical free energy calculationsof a diverse set of HIV-1 protease crystallographic complexeshave enabled a detailed analysis of binding thermodynamics.The thermodynamic consequences of conformational changes thatHIV-1 protease undergoes upon binding to all inhibitors, anda substantial concomitant loss of conformational entropy bythe part of HIV-1 protease that forms the ligand-protein interface,have been examined. The quantitative breakdown of the entropy-drivenchanges occurring during ligand-protein association, such asthe hydrophobic contribution, the conformational entropy termand the entropy loss due to a reduction of rotational and translationsaldegrees of freedom, of a system composed of ligand, proteinand crystallographic water molecules at the ligand-protein interfacehas been carried out The proposed approach provides reasonableestimates of distinctions in binding affinity and gives an insightinto the nature of enthalpy-entropy compensation factors detectedin the binding process.     

Hydrophobic effects on protein/nucleic acid interaction: enhancement of substrate binding by mutating tyrosine 45 to tryptophan in ribonuclease Tl

Hydrophobic effects on binding of ribonuclease Tl to guaninebases of several ribonucleotides have been proved by mutatinga hydrophobic residue at the recognition site and by measuringthe effect on binding. Mutation of a hydrophobic surface residueto a more hydrophobic residue (Tyr45 – Trp) enhances thebinding to ribonucleotides, including mononucleotide inhibitorand product, and a synthetic substrate-analog trinudeotide aswell as the binding to dinucleotide substrates and RNA. Enhancementson binding to non-substrate ribonucleotides by the mutationhave been observed with free energy changes ranging from –2.2 to – 3 .9 kJ/mol. These changes are in good agreementwith that of substrate binding, –2.3 kJ/mol, which iscalculated from Michaelis constants obtained from kinetic studies.It is shown, by comparing the observed and calculated changesin binding free energy with differences in the observed transferfree energy changes of the amino acid side chains from organicsolvents to water, that the enhancement observed on guaninebinding comes from the difference in the hydrophobic effectsof the side chains of tyrosine and tryptophan. Furthermore,a linear relationship between nucleolytic activities and hydrophobicityof the residues (Ala, Phe, Tyr, Trp) at position 45 is observed.The mutation could not change substantially the base specificityof RNase Tl, which exhibits a prime requirement for guaninebases of substrates.     

Gene synthesis and functional expression of a protein exhibiting monodomain IgG Fc binding

A gene encoding a bacterial IgG Fc binding domain was designedand synthesized. The synthetic DNA fragment was cloned 3' toan inducible trpE promoter such that expression of the genein Escherichia coli produced abundant Fc binding protein fusedto the first seven amino acids of the trpE protein. The recombinantprotein contained a single Fc binding domain and demonstratedefficient binding to'human IgG in Western blot analysis. Thisprotein degraded rapidly following cell lysis in the absenceof protease inhibitors, but could be effectively protected bythe addition of protease inhibitor. After purification of theprotein by IgG affinity chromatography, IgG Fc binding abilitywas retained for at least 24 h at either 23 or 37°C andon heating for 15 min at temperatures up to 65°C. No immunoprecipitationwas observed in interactions between the monodomain Fc bindingprotein and IgG molecules. Unlike staphylococcal protein A,no detectable binding of the monodomain IgG Fc binding proteinwas observed to either IgM or IgA. Truncated proteins, expressedfrom a series of 3' deletions of the synthetic gene, were usedto estimate the minimum portion of a monodomain Fc binding proteinthat retained Fc binding ability.     

Fidelity of seryl-tRNA synthetase to binding of natural amino acids from HierDock first principles computations

Seryl-tRNA synthetase (SerRS) charges serine to tRNA(Ser) following the formation of a seryl adenylate intermediate, but the extent to which other non-cognate amino acids compete with serine to bind to SerRS or for the formation of the activated seryl adenylate intermediate is not known. To examine the mechanism of discrimination against non-cognate amino acids, we calculated the relative binding energies of the 20 natural amino acids to SerRS. Starting with the crystal structure of SerRS from Thermus thermophilus with seryl adenylate bound, we used the HierDock and SCREAM (Side-Chain Rotamer Energy Analysis Method) computational methods to predict the binding conformation and binding energy of each of the 20 natural amino acids in the binding site in the best-binding mode and the activating mode. The ordering of the calculated binding energies in the activated mode agrees with kinetic measurements in yeast SerRS that threonine will compete with serine for formation of the activated intermediate while alanine and glycine will not compete significantly. In addition, we predict that asparagine will compete with serine for formation of the activated intermediate. Experiments to check the accuracy of this prediction would be useful in further validating the use of HierDock and SCREAM for designing novel amino acids to incorporate into proteins and for determining mutations in aminoacyl-tRNA synthetase design to facilitate the incorporation of amino acid analogs into proteins.     

Protein stability for single substitution mutants and the extent of local compactness in the denatured state

The stability changes caused by single amino acid substitutionsare studied by a simple, empirical method which takes accountof the free energy change in the compact denatured state aswell as in the native state. The conformational free energyis estimated from effective inter-residue contact energies,as evaluated in our previous study. When this method is applied,with a simple assumption about the compactness of the denaturedstate, for single amino acid replacements at Glu49 of the tryptophansynthase subunit and at Ile3 of bacteriophage T4 lysozyme,the estimates of the unfolding Gibbs free energy changes correlatewell with observed values, especially for hydrophobic aminoacids, and it also yields the same magnitudes of energy as theobserved values for both proteins. When it is also applied foramino acid replacements at various positions to estimate theaverage number of contacts at each position in the denaturedstate from the observed value of unfolding free energy change,those values for replacements with Gly and Ala at the same residueposition in staphylococcal nuclease correlate well with eachother. The estimated numbers of contacts indicate that the proteinis not fully expanded in the denatured state and also that thecompact denatured state may have a substantially native-liketopology, like the molten globule state, in that there is aweak correlation between the estimated average number of contactsat each residue position in the denatured state and the numberof contacts in the native structure. These results provide somefurther evidence that the inter-residue contact energies asapplied here (i) properly reflect actual inter-residue interactionsand (ii) can be considered to be a pairwise hydrophobicity scale.Also, the results indicate that characterization of the denaturedstate is critical to understanding the folding process.     

Rational design and selection of bivalent peptide ligands of thrombin incorporating P4-P1 tetrapeptide sequences: from good substrates to potent inhibitors

The tetrapeptide Phe-Asn-Pro-Arg is a structurally optimized sequence for binding to the active site of thrombin. By conjugating this tetrapeptide or some variants to a C-terminal fragment of hirudin, we were able to generate a series of new bivalent inhibitors of thrombin containing only genetically encodable natural amino acids. We found that synergistic binding to both the active site and an exosite of thrombin can be enhanced through substitutions of amino acid residues at the P3 and P3' sites of the active-site directed sequence, Phe(P4)-Xaa(P3)-Pro(P2)-Arg(P1)-Pro(P1')-Gln(P2')-Yaa(P3'). Complementary to rational design, a phage library was constructed to explore further the residue requirements at the P4, P3 and P3' sites for bivalent and optimized two-site binding. Very significantly, panning of the phage library has led to thrombin-inhibitory peptides possessing strong anti-clotting activities in the low nanomolar range and yet interfering only partially the catalytic active site of thrombin. Modes of action of the newly discovered bivalent inhibitors are rationalized in light of the allosteric properties of thrombin, especially the interplay between the proteolytic action and regulatory binding occurring at thrombin surfaces remote from the catalytic active site.     

Generation and analysis of proline mutants in protein G

The pyrrolidine ring of the amino acid proline reduces the conformational freedom of the protein backbone in its unfolded form and thus enhances protein stability. The strategy of inserting proline into regions of the protein where it does not perturb the structure has been utilized to stabilize many different proteins including enzymes. However, most of these efforts have been based on trial and error, rather than rational design. Here, we try to understand proline's effect on protein stability by introducing proline mutations into various regions of the B1 domain of Streptococcal protein G. We also applied the Optimization of Rotamers By Iterative Techniques computational protein design program, using two different solvation models, to determine the extent to which it could predict the stabilizing and destabilizing effects of prolines. Use of a surface area dependent solvation model resulted in a modest correlation between the experimental free energy of folding and computed energies; on the other hand, use of a Gaussian solvent exclusion model led to significant positive correlation. Including a backbone conformational entropy term to the computational energies increases the statistical significance of the correlation between the experimental stabilities and both solvation models.     

Development of tumor targeting anti-MUC-1 multimer: effects of di-scFv unpaired cysteine location on PEGylation and tumor binding

MUC1 mucin expressed in epithelial cancer, such as prostate and breast, is aberrantly glycosylated providing unique targets for imaging and therapy. In order to create a broadly applicable construct to target these unique epitopes on metastatic cancer, we selected an antibody fragment (scFv) that binds both synthetic MUC1 core peptide and epithelial cancer cell-expressed MUC1, and developed a recombinant bivalent molecule (di-scFv). Genetically engineered modifications of the di-scFv were constructed to create five molecular versions, each having a free cysteine (di-scFv-c) at different locations for site-specific conjugation. The effects of the engineered cysteine in the varied sites were studied relative to tumor binding and polyethylene glycol-maleimide (PEG-Mal) conjugation (PEGylation). Escherichia coli production as well as binding to MUC1 core peptide, human tumor cell lines and human tumor biopsies, were comparable. However, the location of the engineered cysteine in these di-scFv-c did influence PEGylation efficiency of this free thiol; higher PEGylation efficiency occurred with this cysteine in the inter-scFv linkage. Di-scFv-c PEG, with the cysteine engineered after the fifth amino acid in the linker, was used as an example to demonstrate comparable antigen-binding to non-PEGylated di-scFv-c. In summary, novel anti-MUC1 di-scFv-c molecules can be efficiently produced, purified and conjugated by site-specific PEGylation without loss of immunoreactivity, thus providing flexible multidentate constructs for cancer-targeted imaging and therapy.